Spaces:

---

open-llm-leaderboard

/

open_llm_leaderboard

like 14k

Runtime error

Dockerfile

• Updated backend base image to ghcr.io/astral-sh/uv:python3.12-bookworm-slim
• Replaced poetry with uv for dependency installation using uv sync
• Adjusted the startup command to use uv run with uvicorn

backend/app/services/models.py

• Refactored model validation logic, added revision handling for models
• Optimized check for existing submissions by validating model and revision together
• Extracted architectures field from model configuration and formatted it as a string
• Updated model size validation to include revision-specific checks

backend/app/utils/model_validation.py

• Improved model size calculation– integrated safetensors metadata parsing for accurate size validation
• Added revision support for fetching metadata.
• Enhanced error handling for metadata retrieval and model validation